Classification of reviews, error reports and product feature requests using machine learning methods

Abstract

This article proposes a solution for filtering and categorizing user feedback on software products, which can be overwhelming in quantity and often includes uninformative or fake reviews. The proposed approach involves using machine learning methods for classifying reviews into categories such as error reports, product feature requests, and other reviews. The article compares the performance of different classification ML algorithms and investigates the impact of preprocessing options on classification accuracy. Additionally, the article addresses the task of identifying groups of similar reviews in each category, which can be useful for detecting duplicates and identifying patterns. The proposed solution is tested on a dataset and compared with existing solutions. The article concludes by highlighting the novelty and potential benefits of the proposed approach for improving the quality of user feedback and enhancing the reputation of software products.